An intelligent document analysis assistant powered by Gemini AI
Features · Architecture · Tech Stack · Installation · Usage · Deployment · Roadmap
-
Document Intelligence
- Extract key information from PDFs automatically
- Smart summarization of document contents
- Context-aware question answering based on document content
- Support for multiple document formats
-
Advanced UI/UX
- 50-50 split screen display for simultaneous document and chat viewing
- Real-time document analysis and chat interaction
- Responsive design for desktop and mobile devices
- Dark/light mode support
-
Modern Tech Stack
-
Next.js 14 App Router
- Advanced routing for seamless navigation
- React Server Components (RSCs) for optimized rendering
- Server Actions for secure, direct database mutations
-
- Streaming responses for real-time interaction
- Unified API for generating text and structured outputs
- Hooks for building dynamic chat interfaces
-
- Beautifully designed components built with Radix UI and Tailwind
- Fully accessible and customizable interface
- Consistent design language throughout the application
-
-
Security & Performance
- Role-based access control for document sharing
- Document encryption for sensitive files
- Optimized query handling for large documents
- Rate limiting to prevent API abuse
-
Data Persistence
- Vercel Postgres powered by Neon for chat history
- Vercel Blob for document storage
- Document version history and change tracking
-
Authentication
- NextAuth.js for secure user management
- Multiple authentication providers (Google, GitHub, Email)
- JWT token-based session management
erDiagram
User ||--o{ Chat : "has many"
User {
uuid id PK
varchar(64) email
varchar(64) password
}
Chat {
uuid id PK
timestamp createdAt
json messages
uuid userId FK
}
graph TD
User[User] -->|Uploads Document / Asks Question| Frontend[Next.js Frontend]
Frontend -->|Routes Requests| Backend[Next.js Backend/Server]
Backend -->|Document Processing| AI[AI SDK with Gemini]
Backend -->|User Authentication| Auth[NextAuth.js]
Backend -->|Data Operations| Storage[Data Storage]
Storage -->|User & Chat Data| Database[Vercel Postgres]
Storage -->|Document Storage| BlobStore[Vercel Blob]
AI -->|Document Analysis Results| Backend
Auth -->|Authentication Status| Backend
Database -->|Retrieved Data| Backend
BlobStore -->|Retrieved Documents| Backend
Backend -->|Response Data| Frontend
Frontend -->|Display Results| User
graph TD
%% Client Layer
subgraph "Client Layer"
CL1[Web Browser]
CL2[Mobile Device]
end
%% Presentation Layer
subgraph "Presentation Layer"
PL1[Next.js App Router]
PL2[React Components]
PL3[Document Viewer]
PL4[Chat Interface]
PL5[shadcn/ui Components]
end
%% Application Layer
subgraph "Application Layer"
AL1[Next.js Server Actions]
AL2[API Routes]
AL3[Document Processor]
AL4[Chat Controller]
end
%% AI/ML Layer
subgraph "AI/ML Layer"
ML1[AI SDK]
ML2[Gemini Integration]
ML3[Document Analyzer]
ML4[Context Builder]
end
%% Data Access Layer
subgraph "Data Access Layer"
DL1[Authentication Service]
DL2[Database Access]
DL3[Blob Storage Access]
end
%% Storage Layer
subgraph "Storage Layer"
SL1[NextAuth.js]
SL2[Vercel Postgres]
SL3[Vercel Blob]
end
%% Connections
CL1 <--> PL1
CL2 <--> PL1
PL1 --> PL2
PL2 --> PL3
PL2 --> PL4
PL2 --> PL5
PL3 --> AL1
PL4 --> AL1
AL1 --> AL2
AL1 --> AL3
AL1 --> AL4
AL3 --> ML1
AL4 --> ML1
ML1 --> ML2
ML1 --> ML3
ML1 --> ML4
AL1 --> DL1
AL3 --> DL3
AL4 --> DL2
DL1 --> SL1
DL2 --> SL2
DL3 --> SL3
sequenceDiagram
actor User
participant Frontend as Next.js Frontend
participant Backend as Next.js Backend
participant Auth as NextAuth.js
participant DB as Vercel Postgres
participant Blob as Vercel Blob
participant AI as AI SDK (Gemini)
%% Document Upload Flow
User ->> Frontend: Uploads Document
Frontend ->> Backend: Document Upload Request
Backend ->> Auth: Verify User Session
Auth -->> Backend: Session Valid
Backend ->> Blob: Store Document
Blob -->> Backend: Storage Confirmation
Backend ->> DB: Record Document Metadata
DB -->> Backend: Confirmation
Backend -->> Frontend: Upload Success
Frontend -->> User: Display Document
%% Question Answering Flow
User ->> Frontend: Asks Question
Frontend ->> Backend: Process Question
Backend ->> Auth: Verify User Session
Auth -->> Backend: Session Valid
Backend ->> DB: Retrieve Chat History
DB -->> Backend: Chat History
Backend ->> Blob: Retrieve Document
Blob -->> Backend: Document Content
Backend ->> AI: Process Query with Document Context
AI -->> Backend: Generated Response
Backend ->> DB: Save Interaction
DB -->> Backend: Confirmation
Backend -->> Frontend: Response Data
Frontend -->> User: Display Answer
This project provides a database utility module using Drizzle ORM and PostgreSQL to handle users and chats. Below is an overview of the available functions:
| Function | Description |
|---|---|
getUser(email: string) |
Fetches a user by email. |
createUser(email: string, password: string) |
Creates a new user with a hashed password. |
| Function | Description |
|---|---|
saveChat({ id, messages, userId }) |
Saves or updates a chat session. |
deleteChatById({ id }) |
Deletes a chat by its ID. |
getChatsByUserId({ id }) |
Retrieves all chats for a given user, ordered by creation date. |
getChatById({ id }) |
Retrieves a single chat by ID. |
-
Frontend
- Next.js 14 (App Router)
- React 18 with Server Components
- Tailwind CSS for styling
- shadcn/ui component library
- TypeScript for type safety
-
Backend
- Next.js API Routes and Server Actions
- Vercel AI SDK for LLM integration
- Gemini AI for document analysis
- NextAuth.js for authentication
-
Database
- Vercel Postgres (powered by Neon)
- Drizzle ORM for type-safe queries
- Vercel Blob for document storage
-
DevOps
- Vercel for hosting and serverless functions
- GitHub Actions for CI/CD
- pnpm for package management
- Clone the repository:
git clone https://github.com/yourusername/chatpdf.git
cd chatpdf- Install dependencies:
pnpm install- Set up environment variables:
cp .env.example .env- Configure your environment variables in
.envfile:
# Database
POSTGRES_URL=your_postgres_connection_string
# Authentication
NEXTAUTH_URL=http://localhost:3000
NEXTAUTH_SECRET=your_nextauth_secret
GOOGLE_CLIENT_ID=your_google_client_id
GOOGLE_CLIENT_SECRET=your_google_client_secret
# Storage
BLOB_READ_WRITE_TOKEN=your_vercel_blob_token
# AI
AI_API_KEY=your_gemini_api_key
- Start the development server:
pnpm dev-
Access the application at http://localhost:3000
-
Upload a document and start asking questions!
The easiest way to deploy this application is using the Vercel Platform:
pnpm i -g vercel
vercelOr connect your GitHub repository to Vercel for automatic deployments.
For a production build:
pnpm build
pnpm start- Document Chunking: Large documents are automatically chunked into manageable segments for efficient processing
- Response Streaming: AI responses are streamed in real-time for better user experience
- Semantic Caching: Similar questions use cached results to reduce API costs and improve response time
- Progressive Loading: UI components load progressively for faster perceived performance
- Multi-document analysis (comparing information across documents)
- Custom knowledge base integration
- Document annotation and collaborative editing
- Mobile application with offline support
- Webhooks for integration with third-party services
- Advanced document analytics dashboard
This project is licensed under the MIT License - see the LICENSE file for details.
- Vercel for their amazing platform and AI SDK
- Google for the Gemini AI model
- shadcn for the beautiful UI components
- NextAuth.js team for the authentication system
- All contributors who have helped shape this project
Built with ❤️ using Next.js, Gemini, and Vercel